Loss-Of-Function Mutations in ATP6AP1 and ATP6AP2 in Granular Cell Tumors

Thomas Jefferson University Jefferson Digital Commons Department of Pathology, Anatomy, and Cell Department of Pathology, Anatomy, and Cell Biology Faculty Papers Biology

12-1-2018 Loss-of-function mutations in ATP6AP1 and ATP6AP2 in granular cell tumors. Fresia Pareja Memorial Sloan Kettering Cancer Center

Alissa H. Brandes Memorial Sloan Kettering Cancer Center

Thais Basili Memorial Sloan Kettering Cancer Center

Pier Selenica Memorial Sloan Kettering Cancer Center

Felipe C. Geyer Memorial Sloan Kettering Cancer Center

See next page for additional authors

Let us know how access to this document benefits ouy Follow this and additional works at: https://jdc.jefferson.edu/pacbfp Part of the Pathology Commons

Recommended Citation Pareja, Fresia; Brandes, Alissa H.; Basili, Thais; Selenica, Pier; Geyer, Felipe C.; Fan, Dan; Da Cruz Paula, Arnaud; Kumar, Rahul; Brown, David N.; Gularte-Mérida, Rodrigo; Alemar, Barbara; Bi, Rui; Lim, Raymond S.; de Bruijn, Ino; Fujisawa, Sho; Gardner, Rui; Feng, Elvin; Li, Anqi; da Silva, Edaise M.; Lozada, John R.; Blecua, Pedro; Cohen-Gould, Leona; Jungbluth, Achim A.; Rakha, Emad A.; Ellis, Ian O.; Edelweiss, Maria I.A.; Palazzo, Juan P.; Norton, Larry; Hollmann, Travis; Edelweiss, Marcia; Rubin, Brian P.; Weigelt, Britta; and Reis-Filho, Jorge S., "Loss-of-function mutations in ATP6AP1 and ATP6AP2 in granular cell tumors." (2018). Department of Pathology, Anatomy, and Cell Biology Faculty Papers. Paper 241. https://jdc.jefferson.edu/pacbfp/241 Authors Fresia Pareja, Alissa H. Brandes, Thais Basili, Pier Selenica, Felipe C. Geyer, Dan Fan, Arnaud Da Cruz Paula, Rahul Kumar, David N. Brown, Rodrigo Gularte-Mérida, Barbara Alemar, Rui Bi, Raymond S. Lim, Ino de Bruijn, Sho Fujisawa, Rui Gardner, Elvin Feng, Anqi Li, Edaise M. da Silva, John R. Lozada, Pedro Blecua, Leona Cohen-Gould, Achim A. Jungbluth, Emad A. Rakha, Ian O. Ellis, Maria I.A. Edelweiss, Juan P. Palazzo, Larry Norton, Travis Hollmann, Marcia Edelweiss, Brian P. Rubin, Britta Weigelt, and Jorge S. Reis-Filho

This article is available at Jefferson Digital Commons: https://jdc.jefferson.edu/pacbfp/241 ARTICLE

DOI: 10.1038/s41467-018-05886-y OPEN Loss-of-function mutations in ATP6AP1 and ATP6AP2 in granular cell tumors

Fresia Pareja1, Alissa H. Brandes1, Thais Basili1, Pier Selenica1, Felipe C. Geyer1, Dan Fan1, Arnaud Da Cruz Paula1, Rahul Kumar1, David N. Brown1, Rodrigo Gularte-Mérida1, Barbara Alemar1,2, Rui Bi1,3, Raymond S. Lim1, Ino de Bruijn1, Sho Fujisawa4, Rui Gardner5, Elvin Feng4, Anqi Li1,3, Edaise M. da Silva1, John R. Lozada 1, Pedro Blecua6, Leona Cohen-Gould7, Achim A. Jungbluth1, Emad A. Rakha8, Ian O. Ellis8, Maria I.A. Edelweiss9, Juan Palazzo10, Larry Norton11, Travis Hollmann1, Marcia Edelweiss1, Brian P. Rubin12, Britta Weigelt1 & Jorge S. Reis-Filho1 1234567890():,;

Granular cell tumors (GCTs) are rare tumors that can arise in multiple anatomical locations, and are characterized by abundant intracytoplasmic granules. The genetic drivers of GCTs are currently unknown. Here, we apply whole-exome sequencing and targeted sequencing analysis to reveal mutually exclusive, clonal, inactivating somatic mutations in the endosomal pH regulators ATP6AP1 or ATP6AP2 in 72% of GCTs. Silencing of these genes in vitro results in impaired vesicle acidiﬁcation, redistribution of endosomal compartments, and accumulation of intracytoplasmic granules, recapitulating the cardinal phenotypic characteristics of GCTs and providing a novel genotypic–phenotypic correlation. In addition, depletion of ATP6AP1 or ATP6AP2 results in the acquisition of oncogenic properties. Our results demonstrate that inactivating mutations of ATP6AP1 and ATP6AP2 are likely oncogenic drivers of GCTs and underpin the genesis of the intracytoplasmic granules that characterize them, providing a genetic link between endosomal pH regulation and tumorigenesis.

1 Department of Pathology, Memorial Sloan Kettering Cancer Center, New York 10065 NY, USA. 2 Genetics and Molecular Biology, Federal University of Rio Grande do Sul, Porto Alegre 91501-970, Brazil. 3 Department of Pathology, Fudan University Shanghai Cancer Center, Shanghai 200032, China. 4 Molecular Cytology Core Facility, Memorial Sloan Kettering Cancer Center, New York 10065 NY, USA. 5 Flow Cytometry Core Facility, Memorial Sloan Kettering Cancer Center, New York 10065 NY, USA. 6 Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York 10065 NY, USA. 7 Department of Biochemistry, Weill Cornell Medical College, New York 10065 NY, USA. 8 Department of Pathology, University of Nottingham, Nottingham NG7 2RD, UK. 9 Hospital de Clínicas, Federal University of Rio Grande do Sul, Porto Alegre 90035-903, Brazil. 10 Department of Pathology, Jefferson Medical College, Philadelphia 19107 PA, USA. 11 Department of Medicine, Memorial Sloan Kettering Cancer Center, New York 10065 NY, USA. 12 Departments of Pathology and Cancer Biology, Robert J. Tomsich Pathology and Laboratory Medicine Institute and The Lerner Research Institute, Cleveland Clinic, Cleveland 44195 OH, USA. These authors contributed equally: Fresia Pareja, Alissa H. Brandes, Thais Basili. Correspondence and requests for materials should be addressed to B.W. (email: [email protected]) or to J.S.R-F. (email: reisﬁ[email protected])

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y

ranular cell tumors (GCTs) are uncommon neoplasms Our study comprised a discovery cohort (n = 17) and a which can arise in multiple anatomical sites. These validation cohort (n = 65; Fig. 1). To determine whether GCTs G 1 tumors usually follow a benign course , but may occa- would be underpinned by a pathognomonic genetic alteration, we sionally exhibit an aggressive behavior with local and distant subjected the 17 GCTs of the discovery cohort to WES and 11 recurrences1–3. GCTs are characterized by abundant intracyto- GCTs to RNA sequencing, with 10 cases being subjected to both plasmic granules, whose nature and function remain unclear1. (Fig. 1, Supplementary Table 1). These analyses revealed a paucity The genetic landscape of GCTs and the mechanisms under- of CNAs (Supplementary Fig. 1a), no recurrently mutated known pinning the presence of their characteristic intracytoplasmic cancer genes, and no recurrent likely pathogenic fusion genes granules are currently unknown1. (Fig. 2a, Supplementary Table 2). Rather, WES analysis revealed There is a burgeoning body of evidence indicating that genetic recurrent, mutually exclusive, and clonal loss-of-function (i.e., analysis of rare cancer types may provide unique opportunities nonsense or frameshift) somatic mutations affecting ATP6AP1 or for the identification of novel cancer drivers4. A subset of rare ATP6AP2 in 12/17 of the GCTs analyzed (p = 0.041, CoMEt; tumors not uncommonly have simple genomes, with a paucity of Fig. 2a, Supplementary Fig. 1b, Supplementary Tables 3–4). All copy number alterations (CNAs) and somatic mutations, and are ATP6AP1 and ATP6AP2 somatic mutations identified by WES characterized by highly recurrent, specific, or even pathogno- were validated by Sanger sequencing (Supplementary Fig. 1c). To monic, somatic mutations, or fusion genes4. These tumors have validate our findings, we subjected 65 additional GCTs from the distinctive phenotypes and often arise in diverse anatomic loca- validation cohort to targeted massively parallel sequencing, which tions. Akin to these tumors, GCTs are rare, arise in different revealed mutually exclusive loss-of-function mutations (i.e., anatomic locations, and display peculiar morphologic features; nonsense, frameshift, or splice-site) affecting ATP6AP1 and hence, we posited that they could also be underpinned by a highly ATP6AP2 in 36/65 and 6/65 cases, respectively (p = 0.0028, recurrent genetic alteration. CoMEt; Figs. 1 and 2b, Supplementary Table 1). In addition, 5/65 Here, through a whole-exome sequencing (WES) and targeted GCTs harbored ATP6AP1 in-frame indels affecting evolutionarily sequencing analysis of GCTs, we uncovered highly recurrent and conserved residues (Fig. 2b, Supplementary Fig. 1d). All mutually exclusive inactivating mutations targeting the endoso- ATP6AP1 and ATP6AP2 mutations identified by targeted mal pH regulators ATP6AP1 and ATP6AP2 in GCTs. In vitro sequencing (n = 47) were validated by Sanger sequencing and/ silencing of ATP6AP1 and ATP6AP2 in human Schwann cells or repeat targeted capture sequencing analysis (Fig. 1a and and epithelial cells resulted in the accumulation of intracyto- Supplementary Table 1). In total, 72% (59/82) of the GCTs plasmic granules that are ultra-structurally and phenotypically analyzed here harbored likely inactivating somatic mutations in similar to those of human GCTs, altered endosomal acidification ATP6AP1 or ATP6AP2. Of note, no differences in histologic and oncogenic properties, thereby establishing a novel features (Supplementary Table 5) or anatomical site (Fig. 2c) of genotypic–phenotypic correlation. GCTs according to the ATP6AP1/ATP6AP2 mutational status were observed.

Results Recurrent ATP6AP1 and ATP6AP2 somatic mutations in ATP6AP1 and ATP6AP2 inactivating mutations are expressed. GCTs. GCTs were retrieved from the authors’ institutions, fol- ATP6AP1 and ATP6AP2 map to Xq28 and Xp11.4, respectively. lowing the approval of this study by the local research ethics Therefore, a single inactivating mutation in either gene targeting committees or institutional review boards (IRBs) of the con- the X chromosome in males or the active/non-methylated X tributing authors’ institutions. Patient consent was obtained chromosome in females would be sufficient to cause its complete where appropriate, according to the protocols approved. Upon loss of function5. To determine whether ATP6AP1 and ATP6AP2 central pathology review, 82 cases were classified as GCTs, which loss-of-function mutations affect the active/non-methylated X originated in different anatomic locations, including skin (n = chromosome in females, we conducted bisulfite sequencing of 40), soft tissue (n = 15), gastrointestinal tract (n = 13), breast GCTs harboring ATP6AP1 mutations in the vicinity of CpG (n = 6), tongue (n = 5), and other locations (n = 3; Supplemen- islands. No GCTs included in this study harbored ATP6AP2 tary Table 1). mutations adjacent to CpG islands. Bisulfite sequencing revealed

Granular cell tumors and histologic mimics (n = 185)

Granular cell tumors (n = 82) Histologic mimics of granular cell tumors (n = 103) - Paragangliomas (n = 29) - Schwannomas (n = 25) - Adrenocortical carcinomas (n = 16) Discovery cohort (n = 17) Validation cohort (n = 65) - Oncocytomas (n = 13) - Hibernomas (n = 10) Whole-exome sequencing Targeted sequencing - Chromophobe renal cell carcinomas (n = 10) n =7 n =10 n =1 (n =64) RNA sequencing (n = 11)

Validation of ATP6AP1 and Validation of ATP6AP1 and ATP6AP2 mutations by Sanger ATP6AP2 mutations by Sanger sequencing and/or repeat sequencing (n = 12) targeted capture (n = 47) Targeted sequencing (n = 103)

Fig. 1 Schematic representation of the tissue samples and sequencing methods employed in this study. Depiction of the discovery and validation cohorts of granular cell tumors, and the series of histologic mimics of these tumors included in this study, and of the sequencing analysis methods utilized

2 NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y ARTICLE

abValidation cohort (n =65)

Discovery cohort (n = 17) Frame-shift indel ATP6AP1 63% Truncating SNV ATP6AP2 9% Splice site SNV In-frame indel GCT16 GCT92 GCT10 GCT13 GCT15 GCT2 GCT3 GCT5 GCT6 GCT9 GCT1 GCT4 GCT12 GCT14 GCT11 GCT20 GCT8 0 20 40 60 80 Mutated cases (%) ATP6AP1 ATP6AP2 c LMAN1 100 WT TPRXL 80 ATP6AP1/2 MUT FAT3 60 ATP6AP1 ATP6AP2 MUT KCNN3 40 ALS2CL 20

Mutated cases (%) 0 Truncating SNV In-frame indel Benign 3) =6) =5) = Frame-shift indel Missense SNV Atypical = 40) = 15) = 13) n n n n n n Malignant ST ( Skin ( Breast ( Other ( GI tract ( Toungue ( de ATP6AP1/DAPI ATP6AP2/DAPI ) 200 n.s. WT

/ 150 * WT WT T T 100 * * 50 ATP6AP2 ATP6AP1 ATP6AP1 signal/cell ATP6AP1

(% of ATP6AP1/2 0 Methylated N ) * N 200 n.s. WT

MUT T 150

100

* * ATP6AP1 T 50

ATP6AP2 signal/cell ATP6AP2 0 N (% of ATP6AP1/2 N T

MUT 3) Non-methylated = =3) =3) T n n n ( ( ( GCT72 (ATP6AP1 truncating SNV) WT MUT MUT G>A p.W57* 2 ATP6AP2

ATP6AP1/ATP6AP1ATP6AP2

Fig. 2 Inactivating ATP6AP1 and ATP6AP2 somatic mutations are highly prevalent in granular cell tumors. a Recurrent non-synonymous somatic mutations identified in granular cell tumors (GCTs) by whole-exome sequencing (n = 17). Cases are shown in columns and genes in rows. The histologic classification following the Fanburg-Smith criteria is shown in the phenotype bar (top). Mutation types are color-coded according to the legend. SNV, single-nucleotide variation. b Mutation frequencies of ATP6AP1 and ATP6AP2 identified by targeted capture sequencing of additional GCTs of the validation cohort (n = 65). Mutation types are color-coded according to the legend. c Frequency of ATP6AP1 and ATP6AP2 mutations according to anatomical location. The ATP6AP1 and ATP6AP2 mutational status is color-coded according to the legend. GI, gastrointestinal; ST, soft tissue. d Representative Sanger sequencing electropherograms of bisulfite analysis of an ATP6AP1-mutated GCT in a female patient. Arrows highlight the altered nucleotide. *Cytosine converted to thymidine upon bisulfite treatment. e ATP6AP1 and ATP6AP2 protein expression in ATP6AP1-mutated, ATP6AP2-mutated, and ATP6AP1- and ATP6AP2- wild-type GCTs assessed by immunofluorescence (n = 3). Tumor borders are depicted by a dashed line; T, tumor; N, normal stromal/epithelial cells as internal controls. Scale bars, 20 μm. Quantification of ATP6AP1 and ATP6AP2 fluorescent signal/ cell (n = 3; mean ± S.D.); n.s. = non significant, *P < 0.05; two-tailed unpaired t-tests. MUT, mutated; WT, wild type. Experiments in (d, e) were independently performed at least three times that the ATP6AP1 mutations tested were present in non- expression of ATP6AP1- and ATP6AP2-mutant forms was methylated DNA, indicating that these mutations affected the detected in all ATP6AP1- and ATP6AP2-mutant GCTs analyzed active/non-methylated X chromosome of GCTs in females by RNA sequencing and/or complementary DNA (cDNA) (Fig. 2d and Supplementary Fig. 2a). To validate these findings sequencing, respectively (Supplementary Figs. 2c, d). We using an orthogonal approach, we performed a modified human then aimed to determine whether ATP6AP1 and ATP6AP2 androgen receptor (HUMARA) assay following DNA restriction loss-of-function mutations result in reduced protein expression. digestion with the methylation-sensitive restriction enzyme HhaI, As expected, immunofluorescence and western blot which only cleaves non-methylated DNA. These assays showed analyses revealed lower ATP6AP1 and ATP6AP2 protein that mock-digested DNA displayed ATP6AP1 mutations, whereas levels in GCTs harboring ATP6AP1 or ATP6AP2 DNA following treatment with HhaI was wild-type (Supple- mutations, respectively, than in GCTs wild-type for these mentary Fig. 2b). These findings indicate that ATP6AP1 muta- genes and in normal tissues (Fig. 2e and Supplementary Figs. 3a, tions affect the active X chromosome in females. b). Taken together, these data demonstrate that inactivating Next, we sought to determine whether loss-of-function somatic mutations affecting ATP6AP1 and ATP6AP2 are mutations affecting ATP6AP1 and ATP6AP2 result expressed and result in a significant reduction of their respective in their decreased expression. Messenger RNA (mRNA) protein levels.

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y

a b GCT, this study (n = 82) 72% Bladder (n = 129) 1.6% ATP6AP1 (n = 82; GCTs, this study) Mutated cases: 61% Kidney, chromophobe (n =66) 1.5% Mutation type 5 L82Wfs*16/ F274Sfs*12/ Y414Sfs*25/ Stomach (n = 222) 1.4% Trucating SNV X82_splice F274del Y414_S416del 36.6% Lung (n = 1121) 1.3% W427*/ 11% Frame-shift indel Hfs*22 Head & neck (n = 278) 0.7% In-frame indel 0 7.3% Ovarian (n = 557) 0.7% Splice site SNV SP CS TM 6.1%

0.5% of mutations No. AML (n = 200) Missense SNV Colorectal (n = 231) 0.4% 0 100 200 300 400 470 aa TCGA Prostate (n = 333) 0.3% Gene Brain (n = 1101) 0.2% 5 ATP6AP1 (n = 6285; other tumors, TCGA) Mutated cases: 0.27% Breast (n =811) 0.1% ATP6AP1 Kidney, clear cell (n = 499) 0% ATP6AP2 0.24% Uterine (n = 230) 0% 0.02% Thyroid (n = 507) 0% 0 0.02% SP CS TM No. of mutations No. 0204060 80 0 100 200 300 400 470 aa Mutated cases (%) 5 ATP6AP2 (n = 82; GCTs, this study) Mutated cases: 11%

c 6.1% or 80 3.7% 0 1.2% SP 60 MUT CS TM

ATP6AP1 of mutations No. ATP6AP2 MUT 0 100 200 300 350 aa ATP6AP1 40 mutations (%) mutations 5 ATP6AP2 (n = 6285; other tumors, TCGA) Mutated cases: 0.25% 20 R348*/Q 0.21% 0 0.02% Cases with ATP6AP2 ATP6AP2 0 0.03% 25) SPCS TM = 82) =29) = = 16) =13) = 10)

=10) of mutations No. n n n n n n n 0 100 200 300 350 aa GCT ( Truncating SNV Frame-shift indel In-frame indel

Hibernoma ( Splice site SNV Missense SNV Schwannoma (Oncocytoma ( Paraganglioma (

Kidney, chromophobe ( Adrenocortical carcinoma (

Fig. 3 Mutations affecting ATP6AP1 and ATP6AP2 are likely pathognomonic for granular cell tumors. a Frequency and type of ATP6AP1 and ATP6AP2 mutations in 82 granular cell tumors (GCTs) from this study and in 6285 non-hypermutated cancers from The Cancer Genome Atlas (TCGA) including samples from 16 common cancer types. AML, acute myeloid leukemia; SNV, single-nucleotide variant. b Schematic representation of ATP6AP1 and ATP6AP2 protein domains depicting the 59 mutations affecting ATP6AP1 and ATP6AP2 in 82 GCTs from this study and 33 mutations affecting these genes in 6285 non-hypermutated common cancers from TCGA. Mutations are shown on the x-axis, and the frequency of a particular mutation is represented by the height of each 'lollipop' (y-axis). Mutation types are color-coded according to the legend. c ATP6AP1 and ATP6AP2 mutational frequency in 82 GCTs and 103 histologic mimics sequenced in this study. Indel, small insertion and deletion; MUT, mutant; SNV, single-nucleotide variant; WT, wild-type

ATP6AP1/ATP6AP2 mutations are likely pathognomonic for Taken together, these findings provide strong circumstantial GCTs. To determine whether ATP6AP1 and ATP6AP2 loss-of- evidence to demonstrate that ATP6AP1 and ATP6AP2 loss-of- function mutations are pathognomonic for GCTs, we investi- function mutations are pathognomonic for GCTs. gated their presence in 6285 non-hypermutated cancers across 14 common cancer types from The Cancer Genome Atlas (TCGA) studies retrieved from the cBioPortal6.Incontrastto ATP6AP1 and ATP6AP2 inactivation drives the GCT phenotype. ATP6AP1 and ATP6AP2 encode for regulatory subunits the high frequency of likely inactivating ATP6AP1 (61%) and + ATP6AP2 (11%) somatic mutations in the GCTs analyzed in of the vacuolar H -ATPase (V-ATPase) complex, a key regulator of intracellular organelle pH7, and play a role in the control this study, mutations affecting these genes were found in only – 0.27% and 0.25% of common cancers, respectively (Fig. 3a, b). of endosomal acidification and vesicle trafficking8 10. We posited Moreover, at variance with GCTs, where ATP6AP1 and that loss of function of ATP6AP1 and ATP6AP2 would lead to ATP6AP2 mutations were predominantly frameshift or non- altered endocytosis resulting in the presence of the intracyto- sense, consistent with the mutational spectra of potential tumor plasmic granules characteristic of GCTs. Depletion of ATP6AP1 suppressor genes, inactivating mutations targeting ATP6AP1 or ATP6AP2 in primary Schwann cells, a putative cell of origin of and ATP6AP2 were found to be vanishingly rare in common GCTs1, resulted in the accumulation of numerous intracyto- cancers from TCGA, with a prevalence of 0.03% and 0.05% of plasmic structures of heterogeneous morphology with electron- cases, respectively (Fig. 3b). dense and -clear contents, remarkably similar to those observed We next sequenced the entire coding region of ATP6AP1 and in the human GCTs we analyzed (Fig. 4a, b). A similar but less ATP6AP2 in 103 histologic mimics of GCTs, including overt phenotype was observed upon ATP6AP1 or paragangliomas (n = 29), schwannomas (n = 25), adrenocorti- ATP6AP2 silencing in HEK293 cells and MCF-10A cells (Fig. 4b, cal carcinomas (n = 16), oncocytomas (n = 13), hibernomas Supplementary Fig. 4a, b). (n = 10), and chromophobe renal carcinomas (n = 10). These ATP6AP1 and ATP6AP2 are required for endosomal acidifica- are tumors that may occasionally display intracytoplasmic tion8,9, which is essential for proper endocytic flux7. Immuno- granules and may be considered in the differential diagnosis of fluorescence analysis following silencing of either gene GCTs (Fig. 1). This analysis did not reveal any inactivating demonstrated an altered distribution of endosomal compart- somatic mutations affecting either of these genes in the 103 ments, due to accumulation of early endosomes (EEA1-positive) tumors analyzed (Fig. 3c). and recycling endosomes (Rab13-positive), which are endosomal

4 NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y ARTICLE

abSchwann HEK293 siRNA: Human GCTs MUT Control ATP6AP1 ATP6AP1 MUT ATP6AP2 ATP6AP2

Schwann c 400 * *** Schwann HEK293 300 ** EEA1/DAPI Rab13/DAPI LAMP1/DAPI EEA1/DAPI Rab13/DAPI LAMP1/DAPI ** siRNA: 200 ** *** 100 (% relative to control) (% relative No. of positive foci/cell of positive No. Control 0 EEA1 Rab13 LAMP1 *** HEK293 400 **

ATP6AP1 300

200 * *** * * 100 ATP6AP2 (% relative to control) (% relative No. of positive foci/cell of positive No. 0 EEA1 Rab13 LAMP1 siRNA control siRNA ATP6AP1 siRNA ATP6AP2 d Schwann ** 200 ** ** 150 *** * ** shRNA: Control ATP6AP2 100 Control ATP6AP1ATP6AP2 ATP6AP1 50 EEA1 Endosomal marker (% relative to control) (% relative 0 EEA1 Rab13 LAMP1 Rab13 HEK293 250 *** 200 *** HEK293 Schwann LAMP1 ** n.s. 150 ** n.s. 100 shRNA Control Tubulin 50 shRNA ATP6AP1

Endosomal marker shRNA ATP6AP2 (% relative to control) (% relative 0 EEA1 Rab13 LAMP1 compartments with a higher pH11. In addition, silencing of Overall, these data suggest that the characteristic intracytoplasmic ATP6AP1 or ATP6AP2 led to a decreased number of lysosomes granules of GCTs might be due to an abnormal distribution of (LAMP1-positive; Fig. 4c). Consistent with these ﬁndings, we endosomal compartments resulting in accumulation of those with observed increased EEA1 and Rab13 protein levels, coupled with a higher pH. decreased LAMP1 expression upon stable silencing of ATP6AP1 or ATP6AP2 in immortalized Schwann cells and HEK293 cells by Depletion of ATP6AP1 or ATP6AP2 and endosomal acid- western blot analysis (Fig. 4d and Supplementary Fig. 4c). iﬁcation. Regulation of endosomal pH is key for the adequate

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y

Fig. 4 ATP6AP1 and ATP6AP2 loss of function results in a granular phenotype and redistribution of endosomal compartments. a Representative hematoxylin-and-eosin micrographs and transmission electron micrographs of human granular cell tumors (GCTs) harboring ATP6AP1 or ATP6AP2 loss-of- function mutations. Scale bars, 50 μm (left), 1 μm (center), and 0.5 μm (right). b Representative transmission electron micrographs of primary Schwann cells and HEK293 cells transfected with validated short-interfering RNAs (siRNAs) targeting ATP6AP1, ATP6AP2, or non-targeting control. Scale bars, 1 μm (left) and 0.5 μm (right). c Representative confocal micrographs of immunofluorescence analysis of EEA1, Rab13, and LAMP1 (red) and 4-6-diamidino-2- phenylindole (DAPI, blue) expression in primary Schwann cells and HEK293 cells transfected with siRNAs against ATP6AP1, ATP6AP2, or control. Scale bars, 10 μm. Quantification (right) of EEA1-, Rab13-, and LAMP1-positive foci/cell relative to control. Error bars, mean ± S.D. (n = 10). *P < 0.05, **P < 0.01, ***P < 0.001; two-tailed unpaired t-test. d Representative western blot analysis of EEA1, Rab13, and LAMP1 protein levels in immortalized Schwann cells and HEK293 cells with stable knockdown of ATP6AP1, ATP6AP2, or control. Tubulin was used as protein loading control. Quantification (right) of EEA1, Rab13, and LAMP1 protein levels relative to control. Experiments were performed in replicates (Schwann cells, n ≥ 5; HEK293 cells, n ≥ 4). Error bars, mean ± S.D.; n.s., non significant; *P < 0.05, **P < 0.01, ***P < 0.001; two-tailed unpaired t-test

functioning of endocytosis12. Given that ATP6AP1 and domain, which translocates protons across endosomal mem- ATP6AP2 are integral components of the V-ATPase complex7, branes7. We sought to determine whether loss of function of we posited that their loss of function would result in suboptimal ATP6AP1 or ATP6AP2 would result in decreased V-ATPase acidification of endosomal compartments, coupled with an activity by affecting the assembly of the V1 and V0 domains. altered endocytic flux. We sought to determine the effect of Hence, we evaluated the protein levels of ATP6V1A and ATP6AP1 and ATP6AP2 depletion on the acidification of ATP6V0D1, subunits of the V-ATPase V1 and V0 domains, endocytic organelles. Live-cell microscopy experiments using the respectively, in the membrane and cytoplasmic fractions of pH-sensitive pHrodo Red dextran, which displays increased immortalized Schwann cells and HEK293 cells where ATP6AP1 fluorescence with decreasing pH, revealed that transient or stable or ATP6AP2 had been stably silenced, and in control cells. Given silencing of ATP6AP1 or ATP6AP2 in Schwann cells and that the V1 domain is cytoplasmic, membranous ATP6V1A could HEK293 cells resulted in decreased pHrodo Red dextran fluor- be regarded as a surrogate marker for the relative abundance of escence, indicative of reduced acidification of endosomal com- assembled V-ATPase. Silencing of ATP6AP1 or ATP6AP2 partments (Fig. 5a and Supplementary Fig. 4d). Although the resulted in decreased membrane/cytosolic ATP6V1A (Supple- effect on endosomal acidification upon simultaneous transient mentary Fig. 5a), indicating that loss of function of ATP6AP1 or silencing of ATP6AP1 and ATP6AP2 was more pronounced than ATP6AP2 is associated with a defective assembly of the V0 and following silencing of either gene separately (Supplementary V1 domains of the V-ATPase. Interestingly, we also observed that Fig. 4e), the extent of redistribution of endosomal compartments silencing of either gene caused reduced total ATP6V0D1 levels, upon single and double silencing of these genes was comparable suggesting that loss of function of ATP6AP1 or ATP6AP2 leads (Supplementary Fig. 4f). to a decreased biogenesis or stability of the V-ATPase V0 domain We posited that the decreased endosomal acidification (Supplementary Fig. 5a), providing a putative mechanism for the observed following ATP6AP1 or ATP6AP2 silencing would observed decrease in lysosomal acidification upon depletion of be due to abnormal functioning of the V-ATPase. Hence, either gene. Consistent with these findings, we observed a mild we performed a V-ATPase activity assay using the membrane increase in the levels of LC3B-II upon ATP6AP1 or fraction of immortalized Schwann cells and HEK293 cells, where ATP6AP2 silencing in immortalized Schwann cells and ATP6AP1 or ATP6AP2 had been stably silenced, and control HEK293 cells at baseline conditions compared to control cells, in the presence and absence of Bafilomycin-A1 (a selective (Supplementary Fig. 5b), indicating that loss of function of either inhibitor of the V-ATPase). These experiments revealed that gene results in a modest but significant impairment of the silencing of ATP6AP1 or ATP6AP2 led to decreased V-ATPase autophagic flux. activity compared to control (Fig. 5b). To corroborate these Taken together, our findings provide a likely mechanistic basis findings, we evaluated the impact of loss of function of these for the characteristic intracytoplasmic granules of GCTs, given genes on lysosomal activity, as lysosomes are acidic endosomal that ATP6AP1 or ATP6AP2 silencing impairs the assembly of the compartments. Upon transient silencing of ATP6AP1 or V-ATPase and vesicular acidification, resulting in abnormal ATP6AP2 in Schwann cells and HEK293 cells, we observed endocytic flux, which might be accountable for the accumulation decreased lysosomal activity, detected as reduced fluorescence of of endosomal vesicles with a higher pH. the lysosome-specific self-quenched substrate, which acts as endocytic cargo generating fluorescent signal upon lysosomal Oncogenic properties of ATP6AP1 or ATP6AP2 inactivation. degradation. These findings provide further evidence that In light of the high frequency of ATP6AP1 and ATP6AP2 inac- ATP6AP1 and ATP6AP2 loss of function results in abnormal tivating mutations in GCTs, which suggests a putative tumor functioning of the V-ATPase, with a subsequent decrease in suppressor role for these genes, we hypothesized that their loss of lysosomes and lysosomal activity (Fig. 5c and Supplementary function would result in the acquisition of oncogenic properties Fig. 4g). in vitro in GCT-relevant cell models. While stable silencing of We next investigated whether loss of function of ATP6AP1 or either gene in immortalized Schwann cells and HEK293 cells had ATP6AP2 would result in a general alteration of endocytic flux, negligible effects on cellular proliferation in a Cell Titer-Blue through the evaluation of the trafficking of transferrin, which viability assay (Supplementary Fig. 5c), their combined silencing undergoes receptor-mediated endocytosis. We observed that resulted in decreased cellular viability (Supplementary Fig. 5d), transient and stable silencing of ATP6AP1 or ATP6AP2 led to suggesting that synthetic sickness might be the molecular basis for decreased delivery of pHrodo Red transferrin to acidic endosomal the mutual exclusivity of ATP6AP1 and ATP6AP2 mutations in compartments, suggesting that depletion of either gene results in GCTs. Moreover, treatment with N-ethylmaleimide (NEM; an a block of endocytic flux (Supplementary Figs. 4h, i). H+-ATPase inhibitor that blocks vesicular transport) of immor- The activity of the V-ATPase is regulated by several factors, talized Schwann cells and HEK293 cells, where ATP6AP1 or such as the assembly of its cytoplasmic V1 domain, which ATP6AP2 had been stably silenced, led to a significant decrease in hydrolyzes adenosine triphosphate, and its membrane-bound V0 cellular viability compared to controls, suggesting that the

6 NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y ARTICLE

a Schwann HEK293 b siRNA:

Schwann HEK293 150

Control *** *** *** **

100

50 ATP6AP1 Bafilomycin A1-sensitive Bafilomycin

V-ATPase activity (% of control) V-ATPase 0 siRNA:

Control Control ATP6AP1ATP6AP2 ATP6AP1ATP6AP2 ATP6AP2

Schwann HEK293 c Schwann HEK293 150 150 ** 150 ** *** *** *** *** ** ** 100 100 100

50 50 50 MFI (% of control) MFI (% of control) cell area (% of control) Lysosome-specific SQS Lysosome-specific Lysosome-specific SQS Lysosome-specific

pHrodo Red dextran intensity/ pHrodo Red dextran 0 0 0 siRNA: siRNA: siRNA:

Control Control Control Control ATP6AP1ATP6AP2 ATP6AP1ATP6AP2 ATP6AP1ATP6AP2 ATP6AP1ATP6AP2

Fig. 5 ATP6AP1 and ATP6AP2 loss of function leads to decreased acidification of endosomal compartments and V-ATPase activity. a Representative confocal pHrodo Red dextran fluorescence micrographs of primary Schwann cells and HEK293 cells transfected with siRNAs against ATP6AP1, ATP6AP2, or control. Scale bars, 50 μm. Quantification (bottom) of pHrodo Red dextran fluorescence/cell area relative to control. Error bars, mean ± S.D. (n = 3). **P < 0.01, ***P < 0.001; two-tailed unpaired t-test. b Enzymatic V-ATPase activity assay in immortalized Schwann cells and HEK293 cells with stable silencing of ATP6AP1 or ATP6AP2 and control cells, in the presence of Bafilomycin-A1. V-ATPase activity is depicted compared to control. Error bars, mean ± S.D. Schwann cells, n = 4; HEK293 cells, n = 6. **P < 0.01, ***P < 0.001; two-tailed unpaired t-test. c Lysosomal activity assay of primary Schwann cells and HEK293 cells transfected with siRNAs against ATP6AP1, ATP6AP2, or control. Median fluorescence intensity (MFI) of the lysosome-specific self- quenched substrate (SQS) fluorescent signal as determined by flow cytometry, compared to control. Error bars, mean ± S.D. Schwann cells, n ≥ 3; HEK293 cells, n = 3. **P < 0.01, ***P < 0.001; two-tailed unpaired t-test decreased endosomal acidification due to ATP6AP1 or ATP6AP2 and ATP6AP2 loss of function, we assessed the phosphorylation loss of function might render cells susceptible to further V- status of 43 kinases and 2 related proteins upon transient ATPase inhibition (Supplementary Fig. 5e). silencing of ATP6AP1 or ATP6AP2 in primary Schwann cells We observed that stable silencing of ATP6AP1 or ATP6AP2 in and stable silencing of either gene in HEK293 cells and immortalized Schwann cells and HEK293 cells, and transient immortalized Schwann cells. These analyses revealed that silencing of either gene in primary Schwann cells, enhanced depletion of these genes resulted in increased phosphorylation cellular migration in wound healing assays (Fig. 6a and of multiple signaling hubs, such as focal adhesion kinase (FAK) Supplementary Fig. 5f), and resulted in an increased number of and its downstream targets p-38α and Hsp2714, Src-family colonies and average colony size in a soft agar colony formation kinases (SFKs) and signal transducer and activator of transcrip- assay (Fig. 6b and Supplementary Fig. 5g). The increased cellular tion 5a/b (STAT5a/b), 5'-adenosine monophosphate-activated migration and anchorage-independent growth we observed upon protein kinase-α2 (AMPKα2) and its target CREB, which silencing of ATP6AP1 or ATP6AP2 indicates that loss of function facilitates cancer cell adaptation to metabolic stress15, glycogen of either gene results in the acquisition of oncogenic properties synthase kinase-3β (GSK3β), a key component of the Wnt in vitro, supporting a novel tumor suppressor role for ATP6AP1 pathway16, and platelet-derived growth factor receptor and ATP6AP2 and the notion that inactivating mutations β (PDGFR-β) (Fig. 6c). To determine whether the oncogenic targeting these genes are potential drivers of GCTs. properties elicited by ATP6AP1 or ATP6AP2 loss of function Endocytosis orchestrates an array of cellular processes, may be attributed to increased signaling via the pathways above, including signaling, as it fine-tunes the amplitude and duration we inhibited them pharmacologically, and assessed whether the of various signaling cascades13. Hence, we posited that altered oncogenic phenotype observed due to silencing of these genes endocytosis downstream of ATP6AP1 or ATP6AP2 loss of would be reversed. Treatment with PD166285 (a protein tyrosine function would result in the activation of signaling hubs. To kinase inhibitor which potently targets PDGFR-β17), PP2 (a identify the signaling cascades that may be triggered by ATP6AP1 selective inhibitor of SFKs (which inhibits Lck, Fyn, and Hck

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y

ab Schwann HEK293 shRNA: Schwann HEK293 shRNA: Control Control #1 ATP6AP1 #2 ATP6AP1 #1 ATP6AP2 ATP6AP2 #2 #1 #2 #1 #2 0 h 16 h 0 h 16 h

Schwann HEK293 Schwann HEK293 100 300 *** ** * 80 *** ** ** ** *** ** * * ** ** ** 60 *** 200 n.s.

(area) 40 100

20 (% of control) % Wound % Wound healed

0 of colonies/well No. 0 shRNA: shRNA:

Control Control Control Control

ATP6AP1ATP6AP1 #1ATP6AP2 #2ATP6AP2 #1 #2 ATP6AP1ATP6AP1 #1ATP6AP2 #2ATP6AP2 #1 #2 ATP6AP1ATP6AP1 #1ATP6AP2 #2ATP6AP2 #1 #2 ATP6AP1ATP6AP1 #1ATP6AP2 #2ATP6AP2 #1 #2

c d

Schwann 250 STAT5a (Y694) FAK (Y397) 200 * EGFR (Y1086) * ** Yes (Y426) 150 ** β-Catenin * CREB (S133) 100 * MSK1/2 (S376/S360) (% of control) PYK2 (Y402) 50 RSK1/2/3 (S380/S386/S377) of colonies/well No. STAT2 (Y689) 0 STAT6 (Y641) Control PD166285 PP2 CAS GSK-3α/β (S21/S9) (DMSO) (50 nM) (500 nM) (50 μM) AMPKα2 (T172) Fyn (Y420) HSP27 (S78/S82) Hck (Y411) PRAS40 (T246) c-Jun (S63) Fold change p53 (S46) 3 TOR (S2448) 2 Akt1/2/3 (T308) 1 JNK1/2/3 (T183/Y185, T221/Y223) 0.7 ERK1/2 (T202/Y204, T185/Y187) eNOS (S1177) 0.5 p53 (S392) P < 0.05, t-test p70 S6 Kinase (T389) AMPKα1 (T183) Cell model STAT3 (Y705) STAT5a/b (Y694/Y699) Primary Schwann STAT5b (Y699) Immortalized Schwann p38α (T180/Y182) HEK293 PDGFRβ (Y751) Lyn (Y397) Gene silenced Fgr (Y412) Chk-2 (T68) ATP6AP1 p27 (T198) ATP6AP2 robustly18), and the STAT5 inhibitor CAS 285986-31-419 with PD166285 and PP2 blocked the increased cellular migration partially reversed the increased anchorage-independent growth observed upon depletion of ATP6AP1 or ATP6AP2 (Supple- elicited by stable silencing of ATP6AP1 or ATP6AP2 in mentary Fig. 5h). These results demonstrate that the oncogenic immortalized Schwann cells (Fig. 6d). In addition, treatment properties downstream of ATP6AP1 or ATP6AP2

8 NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y ARTICLE

Fig. 6 Loss of function of ATP6AP1 and ATP6AP2 confers oncogenic properties in vitro. a Wound healing assay of immortalized Schwann cells and HEK293 cells with stable silencing of ATP6AP1 or ATP6AP2 using short-hairpin RNAs (shRNAs). Wound area was assessed at 0 and 16 h. Scale bars, 500 μm. Quantification (bottom) of wound healed compared to control. Error bars, mean ± S.D. Schwann cells, n = 6; HEK293 cells, n = 3. *P < 0.05, **P < 0.01, ***P < 0.001; two-tailed unpaired t-test. b Soft agar colony formation assay of immortalized Schwann cells and HEK293 cells with stable shRNA silencing of ATP6AP1, ATP6AP2, or control shRNAs. Scale bars, 500 μm. Quantification (bottom) of number of colonies/well compared to control. Error bars, mean ± S.D. Schwann cells, n = 4; HEK293 cells, n = 6. n.s., non significant; *P < 0.05, **P < 0.01; two-tailed unpaired t-test. c Heatmap depicting phosphorylation levels of selected target proteins upon transient (primary Schwann) or stable (immortalized Schwann and HEK293 cells) RNA-interference silencing of ATP6AP1 or ATP6AP2, relative to control. Protein phosphorylation fold change, cell model, and gene silenced are color-coded according to the legend. Only proteins significantly (P < 0.05; two-tailed unpaired t-test) altered in at least one cell model are shown. Experiments were performed in replicates (primary Schwann cells, n = 4; immortalized Schwann cells, n = 4; HEK293 cells, n = 4). d Soft agar colony formation assay of immortalized Schwann cells with stable silencing of ATP6AP1, ATP6AP2, or control following 14 days of treatment with 50 nM of PD166285, 500 nM of PP2, and 50 μM of CAS 285986-31-4 (CAS) or vehicle (DMSO). Error bars, mean ± S.D. (n >5)*P < 0.05, **P < 0.01; two-tailed unpaired t-test

loss of function might be mediated, at least to some extent, by ATPase V0 domain levels in floxed mouse embryonal fibroblasts increased signaling via PDGFR-β, SFKs, and STAT5 pathways. following Atp6ap2 deletion. Notably, in yeast, the loss of any V0 subunit may affect the stability of the remaining V0 subunits, Discussion and the assembly of the yeast V-ATPase requires assembly fac- 29 Here, we demonstrate that GCTs harbor recurrent, mutually tors, such as Vma12p, Vma21p, and Vma22p , and mutant cells exclusive, and clonal somatic loss-of-function mutations affecting lacking these factors display low levels of the V-ATPase V0 30 ATP6AP1 and ATP6AP2. These genes encode for accessory domain . One could posit that ATP6AP1 and ATP6AP2 activity proteins of the V-ATPase, which plays a pivotal role in the reg- may be required for the assembly of the V-ATPase in humans, ulation of endosomal pH7. Missense germline mutations of and that their loss of function results in decreased stability of the ATP6AP1 have been shown to be causative of Immunodeficiency V0 domain with a subsequent reduction of V-ATPase functional 4720, characterized by hepatopathy, cognitive impairment, and levels. abnormal protein glycosylation, whereas missense germline Our functional studies demonstrated that ATP6AP1 and mutations affecting ATP6AP2 result in the Hedera-type X-linked ATP6AP2 silencing results in the acquisition of oncogenic mental retardation syndrome21,22. Importantly, however, the role properties in vitro, which are partially dependent on signaling via of loss-of-function germline mutations affecting ATP6AP1 and PDGFR-β, SFKs, and STAT5. Endocytosis and cell signaling are 31 ATP6AP2 remains to be determined and neither ATP6AP1 nor intimately related cellular processes , and it is therefore not ATP6AP2 have been previously implicated in cancer. We and surprising that loss of function of ATP6AP1 and ATP6AP2 leads others have shown that a subset of rare tumors arising in multiple to the activation of several signaling pathways. It should be noted, anatomic sites are not uncommonly underpinned by recurrent, however, that defective endocytosis downstream of inactivation of specific, or even pathognomonic, genetic alterations23,24. In this either of these genes is possibly the common underlying study, we expand the spectrum of rare cancer types underpinned mechanistic basis resulting in activation of different signaling by likely pathognomonic genetic alterations, as loss-of-function pathways. Further studies are warranted to define the mechanistic mutations targeting ATP6AP1 or ATP6AP2 are present in up to links between ATP6AP1 and ATP6AP2 loss of function and 72% of GCTs, but are found in less than 0.1% of common cancer altered signaling. types (Fig. 3a, b) and in none of the histologic mimics of GCTs Our study has several limitations. Due to the multi- tested here (Fig. 3c). institutional nature of our study, follow-up information was The histogenesis of GCTs remains a matter of contention. unavailable, and an assessment of the impact of ATP6AP1 and While it has been suggested that GCTs derive from Schwann ATP6AP2 mutational status on patient outcome could not be cells1, other cell lineages have also been proposed25–27. Due to the performed. It is possible that altered endocytosis with accumu- lack of representative cell line models derived from human GCTs, lation of endosomal organelles due to a deficient V-ATPase and we established cell models using Schwann cells, the likeliest cell of the oncogenic phenotype downstream of overt signaling via origin of GCTs, and HEK293 cells, where we stably silenced PDGFR-β, SFKs, and STAT5 are phenomena due to loss of ATP6AP1 and ATP6AP2. The in vitro studies we conducted function of ATP6AP1 or ATP6AP2, but independent of each revealed that depletion of ATP6AP1 or ATP6AP2 results in other. accumulation of intracytoplasmic granules, recapitulating the In conclusion, ATP6AP1 and ATP6AP2 loss-of-function cardinal histologic and ultra-structural features of human GCTs mutations are the likely drivers of GCTs, and appear to be (Fig. 4a, b), likely due to accumulation of endosomal compart- pathognomonic for these tumors. Our findings provide a ments with higher pH. These results support a genotypic–phenotypic correlation and indicate that the intracy- genotypic–phenotypic correlation between loss-of-function toplasmic granules of GCTs may constitute accumulated high-pH mutations affecting ATP6AP1 and ATP6AP2 and GCTs. Inter- cytoplasmic vesicles resulting from defects in vesicle acidification. estingly, the accumulation of intracytoplasmic granules was more Finally, by studying a rare tumor type we have uncovered a novel conspicuous in Schwann cells than in epithelial cells, suggesting potential tumor suppressor role for genes essential to endosomal that the interplay between loss of function of ATP6AP1 and pH control and vesicular trafficking. ATP6AP2 and cell of origin is a likely determinant of the novel – genotypic phenotypic correlation described here. Methods We observed that loss of function of ATP6AP1 and ATP6AP2 Subjects and samples. Following approval by the IRB of the authors’ institutions, results in decreased V-ATPase activity and endosomal acidifica- frozen and/or formalin-fixed paraffin-embedded (FFPE) tissue blocks of GCTs tion, likely due to reduced levels of the V0 domain of the V- were retrieved from the pathology archives/tissue banks of Memorial Sloan Ket- tering Cancer Center (MSKCC; NY, USA), Cleveland Clinic (OH, USA), University ATPase, and decreased assembly of the V0 and V1 V-ATPase of Nottingham (UK), Thomas Jefferson University Hospital (PA, USA), and domains. The reduced levels of V-ATPase V0 domain are in Edelweiss Laboratory (Brazil). Patient consents were obtained if required by IRB agreement with Kinouchi et al.28, who reported on decreased V- protocols approved by the authors’ institutions. Samples were anonymized before

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y tissue processing. All cases were centrally reviewed by four pathologists (F.P., F.C. (for mock-digested control). PCR fragments were cleaned using ExoSAP-IT G., T.H., and M.E.) and assessed following the Fanburg-Smith criteria32. In total, (ThermoFisher Scientific) and subjected to Sanger sequencing as previously after central pathology review, 82 tumors (5 frozen, 77 FFPE) arising from different described23. anatomic locations, excluding the central nervous system, were classified by the four study pathologists as GCTs and included in this study. Expression of mutant ATP6AP1 and ATP6AP2. The mRNA expression of mutant The discovery series comprised 17 GCTs (Fig. 1, Supplementary Table 1). DNA ATP6AP1 and ATP6AP2 in GCTs identified by WES was assessed in the cases was extracted separately from microdissected tumor and matched normal tissue subjected to RNA sequencing by visual inspection of reads in IGV47. In addition, samples under a stereomicroscope and subjected to WES. RNA was extracted from the expression of ATP6AP1 and ATP6AP2 mutations in GCTs identified by tar- 11 microdissected GCTs, 10 of which were also subjected to WES, and subjected to geted sequencing was assessed by Sanger sequencing of RNA-derived cDNA. PCR RNA sequencing. The validation series consisted of 65 GCTs; FFPE-derived DNA fragments were cleaned using ExoSAP-IT (ThermoFisher Scientific) and Sanger from microdissected tumor and matched normal tissue samples was subjected to sequenced as previously described23. massively parallel sequencing using a custom bait set targeting all exons and flanking intronic regions of ATP6AP1 and ATP6AP2 (Fig. 1, Supplementary Table 1). In addition, DNA extracted from microdissected tumor and normal tissue Mutual exclusivity test. Mutual exclusivity analysis of mutations targeting samples from 103 GCT histologic mimics (Fig. 1) was subjected to ATP6AP1 and ATP6AP1 and ATP6AP2 was performed using CoMEt57, a statistical approach to ATP6AP2 targeted massively parallel sequencing. For power calculations, we identify combinations of mutually exclusive alterations in cancer, as previously assumed that GCTs would be driven by a highly recurrent mutation or fusion gene described58. present in ≥70% of cases, akin to other rare tumor types33,34. Based on a binomial distribution, the analysis of 10 cases would be sufficient to identify a highly recurrent genetic alteration or a highly recurrently altered gene with >90% Amino acid sequence alignment of ATP6AP1 across species. To determine statistical power. whether ATP6AP1 residues affected by in-frame indels are evolutionarily conserved, the amino acid sequences of the ATP6AP1 protein from different species were retrieved from Ensembl and aligned using Clustal Omega at EMBL-EBI59,as RNA sequencing and fusion gene identification. RNA sequencing was per- previously described23. formed on 11 GCTs using validated protocols23,35 employed at MSKCC Integrated Genomics Operation (IGO). In brief, paired-end RNA sequencing was performed with 2 × 50 bp cycles on an Illumina HiSeq2000. Read pairs supporting fusion Analysis of data from TCGA. ATP6AP1 and ATP6AP2 mutation frequencies in fi 36 37 6285 common cancer (non-GCT) samples from TCGA were obtained from transcripts were identi ed using INTEGRATE and deFuse , as previously 6 described23,35. Candidates supported by at least two spanning reads were included. cBioPortal , including 14 studies (see Supplementary methods). Following the approach described by Jelinic et al.60, ATP6AP1 and ATP6AP2 mutation fre- To account for alignment artifacts and normal transcriptional variants, we exclu- fi ded fusion gene and read-through candidates which were identified in a set of 38 quencies were assessed following exclusion of hypermutated cases, de ned as normal samples from TCGA38. The remaining candidate fusion genes were cancers harboring more than 1000 non-synonymous mutations, microsatellite 39 fi unstable or harboring POLE or POLD1 exonuclease domain mutations. Mutation annotated using OncoFuse to de ne their likelihood of constituting potential ‘ ’ 6 driver fusion genes. diagrams ( lollipop plots) were generated using MutationMapper on cBioportal and manually curated.

Whole-exome sequencing and variant calling. Microdissected tumor and nor- Immunofluorescence mal DNA samples from 17 GCTs were subjected to WES at MSKCC IGO using . HEK293 (20,000 cells/well) and primary Schwann cells 23,40 (7500 cells/well) were plated on Millicell EZ SLIDE 8-chamber slides, fixed with 4% validated protocols as previously described . Sequencing data were analyzed as fi previously described40. In brief, reads were aligned to the reference human genome formaldehyde (ThermoFisher Scienti c), and stained using primary antibodies GRCh37 using the Burrows-Wheeler Aligner (BWA, v0.7.10)41. Local realignment, against EEA1 (Clone C45B10; Cell Signaling; #3288), Rab13 (Sigma-Aldrich; #SAB4200058), and LAMP1 (Clone D2D11; Cell Signaling; #9091S), followed by duplicate removal, and base quality recalibration were performed using the Gen- – fi ome Analysis Toolkit (GATK, v3.1.1)42. Somatic single-nucleotide variants (SNVs) Alexa Fluor conjugated secondary antibodies (ThermoFisher Scienti c), at 43 recommended dilutions. Slides were mounted using ProLong Gold Antifade were detected by MuTect (v1.0) , and small insertion and deletions (indels) by fi Strelka (v2.0.15)44 and VarScan2 (v2.3.7)45. SNVs and indels for which the tumor Reagent with 4-6-diamidino-2-phenylindole (DAPI; ThermoFisher Scienti c). Formalin-fixed cell pellets were generated, as previously described61. mutant allele fraction (MAF) was <5 times that of the matched normal MAF were fl excluded46. SNVs and indels found at >5% global minor allele frequency in dbSNP Immuno uorescence staining of FFPE tumor and cell pellet sections was (build 137) were also excluded. All mutations were manually inspected using the performed using the Leica Bond RX automated stainer (Leica Biosystems) using 47 48 primary antibodies against ATP6AP1 (OriGene; #TA590072; 1:150) or ATP6AP2 Integrative Genomics Viewer (IGV) . FACETS was used to determine CNAs TM and whether genes harboring a somatic mutation were targeted by loss of het- (Sigma; #HPA003156; 1:200), followed by Alexa Fluor 488 Tyramide Reagent 40,46 49 (ThermoFisher Scientific; #B40953) and DAPI solution (ThermoFisher Scientific), erozygosity, as previously described . ABSOLUTE (v1.0.6) was employed to ’ determine the cancer cell fraction (CCF) of each mutation, as previously descri- according to the manufacturer s instructions. bed40,46. A mutation was classified as clonal if its probability of being clonal was Fluorescence images were acquired with an Axio Imager 2 upright microscope >50%50 or if the lower bound of the 95% confidence interval of its CCF was >90% (Zeiss) with a 40×/0.75 air objective, an Axiocam 506 camera (Zeiss), and Zen Blue 40,46. Mutations affecting hotspot codons51 were annotated as previously (version 2) acquisition software (Zeiss) at a resolution of 1376 × 1104 pixels, a described46. scaling of 0.227 micron/pixel, and a bit depth of 14-bit. DAPI and Alexa488 were illuminated with an Illuminator HXP 120 V(D) lamp (Zeiss) and imaged in the range of 435–485 nm and 512–542 nm, respectively. Linear LUT was used at full Validation of ATP6AP1 and ATP6AP2 mutations. DNA from GCTs of the range. No post-acquisition processing was done, besides minor adjustments of validation series and DNA from histologic mimics of GCTs were subjected to brightness and contrast, applied equally to all images. ImageJ software was used to targeted capture massively parallel sequencing using a custom bait set targeting all quantify the signal intensity per cell; at least five representative images (40× field) exons and flanking intronic regions of ATP6AP1 and ATP6AP2 (IDT Technolo- were analyzed for each case. gies). Sequencing read alignment and local realignment, duplicate removal, and Confocal images were acquired with a TCS SP5 Upright Confocal microscope quality score recalibration were performed using BWA41 and GATK42 as described (Leica Microsystems), using a 63×/1.40 oil objective, HyD hybrid detectors, and the above. SNVs were identified using MuTect43; indels were identified using Strelka44, Leica Application Suite Advanced Fluorescence (LASAF) acquisition software VarScan245, Platypus52, and Scalpel53, and further curated by manual inspection (Leica Microsystems), as previously described62. Images were acquired at a using IGV47. resolution of 1024 × 1024 pixels, a scaling of 0.12 micron/pixel, and an 8-bit depth. All somatic ATP6AP1 and ATP6AP2 mutations identified by WES in GCTs DAPI, Alexa568 and Alexa488 were excited with 405 nm, 543 nm, and 488 nm from the discovery series were validated by Sanger sequencing, as previously lasers, respectively, and imaged in the range of 410–460 nm, 570–620 nm, and described 23. In addition, all somatic ATP6AP1 and ATP6AP2 mutations identified 490–520 nm, respectively. Linear LUT was used at full range. No post-acquisition by targeted sequencing in GCTs from the validation series were validated by Sanger processing was performed, besides minor adjustments of brightness and contrast, sequencing 23 and/or by repeated targeted capture massively parallel sequencing applied equally to all images. ImageJ was used to quantify the number of foci per using an independent DNA sample utilizing our custom ATP6AP1 or ATP6AP2 cell, with at least 7 representative images analyzed for each condition. All bait set (Supplementary Table1). experiments were repeated three times.

Bisulfite sequencing and modified HUMARA assay. For bisulfite sequencing54 Cell lines. HEK293 (ATCC), MCF-10A (ATCC), primary Schwann (ScienCell), of ATP6AP1 mutations near CpG islands, 200 ng of tumor DNA was treated with a and immortalized human normal Schwann cells (Margaret Wallace, University of bisulfite conversion kit (Methylamp DNA Modification Kit; EpiGentek), and Florida)63 were authenticated using short tandem repeat profiling at MSKCC IGO amplified by PCR with methylated and non-methylated DNA-specific primers. For and tested for mycoplasma using the PCR-based Universal Mycoplasma Detection the modified HUMARA assay55 following methylation-specific DNA digestion56, kit (ATCC). HEK293 and immortalized Schwann cells were cultured in Dulbecco's 200 ng of tumor DNA was incubated overnight at 37 °C with 25 U of the modified Eagle's medium (DMEM) high glucose supplemented with 10% fetal methylation-sensitive restriction enzyme HhaI (New England Biolabs) or water bovine serum (FBS) and 1% penicillin/streptomycin. MCF-10A cells were cultured

10 NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y ARTICLE in DMEM/F12 supplemented with 5% horse serum, 20 ng/ml epidermal growth guidelines. Confocal fluorescence and bright-field z-stacks were acquired at 37 °C factor, 10 μg/ml insulin, 0.5 μg/ml hydrocortisone, and 1% penicillin/streptomycin. using an LSM880 inverted confocal microscope (Zeiss) with a 25× objective, Primary Schwann cells were cultured in Schwann Cell Medium (ScienCell), con- Gallium arsenide phosphide (GaAsP) array detectors, and ZEN Black (Version 2.3) sisting of basal medium, 5% FBS, 1% Schwann cell growth supplement, and 1% acquisition software (Zeiss). Images were acquired at a resolution of 1024 × 1024 penicillin/streptomycin. All cell lines were maintained in a 5% CO2 atmosphere at pixels, a scaling of 0.21 × 0.21 × 0.6 micron/pixel, and an 8-bit depth. pHrodo Red 37 °C. dextran was excited with a 561 nm laser and imaged in the range of 562–633 nm. Linear LUT was used at full range. No post-acquisition processing was performed, besides minor adjustments of brightness and contrast, applied equally to all images. Inhibitors. The ATPase inhibitor NEM (#34115, Millipore Sigma) was used in cell Sum-intensity projection images were generated and total pHrodo Red dextran viability assays. The V-ATPase inhibitor Bafilomycin-A1 (Millipore Sigma; intensity per cell-covered area was quantified using ImageJ. Experiments were #B1793) was used in LC3B western blots. PP2 (R&D Systems; #1407), an inhibitor performed in triplicate. of the Src family kinases, PD166285 (#3785, R&D Systems), a Src, PDGFRβ and FGFR1 inhibitor, and CAS 285986-31-4 (Millipore Sigma; #573108), a STAT5 inhibitor, were used in colony formation and wound healing assays (see below). Flow cytometry. For endocytosis experiments, cells were incubated with 25 µg/ml pHrodo Red Transferrin conjugate (ThermoFisher Scientific; #P35376) at 37 °C for 20 min. Fluorescence was evaluated by flow cytometry using an LSR Fortessa (BD Transient siRNA transfection. Transient transfections were performed using Biosciences) instrument with a 561 nm laser and a 582/15 nm BP for excitation and Lipofectamine RNAiMAX (ThermoFisher Scientific) with ON-TARGETplus detection of pHrodo, respectively. The cell population was first gated in a FSC-A vs. SMART pool short-interfering RNA (siRNA) for human ATP6AP1 and ATP6AP2, SSC-A plot, followed by doublet discrimination using FSC-H vs. FSC-W. For single ON-TARGETplus siRNAs for human ATP6AP1 and ATP6AP2, and ON- experiments with GFP+ cells, the GFP+ population was gated using a 2D plot of TARGETplus Non-Targeting control pool (GE Healthcare Dharmacon), according 582/15 nm (561 nm excitation) vs. 530/50 nm (488 nm excitation). In all samples, a to the manufacturer’s instructions. Transfection efficiency was assessed by western histogram of pHrodo fluorescence was measured from the final gated population blotting (see below) and by TaqMan quantitative reverse transcription-polymerase (singlets or GFP+; Supplementary Fig. 8a). chain reaction (qRT-PCR); ATP6AP1 (Hs00184593_m1) and ATP6AP2 For lysosomal activity experiments, the Lysosomal Intracellular Activity Assay (Hs00997145_m1) gene expression levels were evaluated, using GAPDH Kit (Biovision; #K448) was used according to the manufacturer’s instructions. Cells (Hs02786624) for normalization, as previously described64. Experiments were were incubated with the lysosome-specific self-quenched substrate at 37 °C for 1 h. performed 72 h after transfection and repeated a minimum of three times for each The lysosome-specific self-quenched substrate fluorescent signal was detected by condition. flow cytometry using a 3-laser (405 nm, 488 nm; 640 nm) Aurora (Cytek Biosciences) spectral analyzer with 38 fluorescent channels and 2 scatter detectors Generation of stable cell lines. Custom high efficiency miR-E short-hairpin (FSC and SSC). Raw spectral data were acquired and unmixed using Spectroflo RNAs (shRNAs) in the SGEP vector (puromycin/GFP) targeting ATP6AP1 or Software (Cytek Biosciences) with autofluorescence extraction. The cell population ATP6AP2 were created by the MSKCC RNAi Core Facility, as previously descri- was first gated in a FSC-A (488 nm) vs. SSC-A (405 nm) plot, followed by doublet bed65. miR-E shRNA targeted to Firefly Renilla Luciferase was used as a negative, discrimination using FSC-H vs. FSC-W (Supplementary Fig. 8b). Experiments were non-targeting control. Virus containing shRNA was produced in HEK293T cells performed in triplicate and 10,000 events were acquired per sample. The MFI of (ATCC) using a lentiviral packaging mix (Sigma-Aldrich). In addition, Mission each sample was evaluated using FCS express 6.04 (De Novo Software). shRNA lentiviral transduction particles (Sigma-Aldrich) targeting ATP6AP1 or ATP6AP2 or non-targeting control were employed. HEK293 or immortalized V-ATPase activity assay. Total ATPase activity was measured in cell membrane Schwann cells were infected for 48 h with virus containing shRNA and then protein using the ATPase/GTPase Activity Assay Kit (Sigma-Alrich; #MAK-113), selected for 14 days in puromycin (2 µg/ml;ThermoFisher Scientific). Silencing according to the manufacturer’s instructions, in either the presence or absence of efficiency was confirmed by qRT-PCR and western blot, and for ATP6AP1 or 250 nM Bafilomycin-A1. Absorbance at 595 nm was detected using a Victor X4 ATP6AP2, the two shRNAs with the highest level of silencing were selected for Multimode Plate Reader (PerkinElmer). The Bafilomycin-A1-sensitive ATPase downstream experiments. activity was taken as a measure of V-ATPase activity. Experiments were performed in triplicate. Transmission electron microscopy. Cells were fixed with a modified Karnovsky’s fixative, and embedded in an epon analog resin. En face ultrathin sections (65 nm) Scratch wound healing assay. Cells were serum-starved and seeded in 24-well were contrasted with lead citrate, as previously described66. FFPE tissue punches plates at 90–95% confluence in the CytoSelect 24-well Wound Healing Assay(Cell were reprocessed and embedded in flat molds and processed following the above Biolabs, Inc), according to the manufacturer’s guidelines. Phase-contrast images protocol66. were obtained at 0 h and 16 h following scratch wounding, using an EVOS XL Core Transmission electron micrographs were captured using a JSM 1400 electron Microscope (ThermoFisher Scientific) and analyzed using ImageJ. Experiments microscope (JEOL), a Veleta 2K x 2K CCD camera (EMSIS), and the iTEM were performed in triplicate at least 3 times. acquisition software (EMSIS). Images were acquired at a resolution of 4008 × 2672 pixels at 9 micron/pixel and a 14-bit depth. Linear LUT was used at full range. No post-acquisition processing was performed, besides minor adjustments of Colony formation assay. Soft agar colony formation assay was performed using brightness and contrast, applied equally to all images. At least 10 representative CytoSelect 96-well Cell Transformation Assay (Cell Biolabs, Inc.), according to the images (8000× field) were analyzed per condition using ImageJ. manufacturer’s guidelines. Briefly, cells were seeded in 1.2% agar in 96-well plates (1000 cells/well); n = 4 for immortalized Schwann cells and n = 6 for HEK293 cells). After 14 days, phase-contrast images were obtained using the EVOS XL Core Protein blotting. Standard western blotting was conducted as previously descri- Microscope (ThermoFisher Scientific). Quantification of the number of colonies bed67. Frozen tissue was lysed in RIPA buffer and homogenized using a Tissuelyser per well and colony size was performed using ImageJ as previously described23. (Qiagen). Proteins were extracted from FFPE material using the Qproteome FFPE Tissue Kit (Qiagen; #37623), according to the manufacturer’s instructions. Cell fractionation was performed using the Cell Fractionation Kit (Cell Signaling; Proliferation assay. Cells were seeded in 96-well plates (1000 cells/well; n = 3 for #9038), according to the manufacturer’s instructions. Primary antibodies against immortalized Schwann cells and n = 4 for HEK293 cells). Proliferation rate was ATP6AP1 (cell lines and cell pellets: Santa Cruz; #sc-81886; GCTs: Sigma-Aldrich; assessed using the Cell Titer-Blue Cell Viability Assay (Promega). Absorbance #A1486), ATP6AP2 (Sigma-Aldrich; #HPA003156), Tubulin (Cell Signaling; detection was performed with 560 nm excitation and 590 nm emission using a #2125), GAPDH (Clone 14C10; Cell Signaling; #2118), ATP6V1A (Abcam; Victor X4 Multimode Plate Reader (PerkinElmer), as previously described. #ab137574), ATP6V0D1 (Abcam; #ab56441), MEK1/2 (Clone D1A5; Cell Signal- Experiments were performed in triplicate. ing; #8727), NRG1 (Abcam; #ab180808), LC3B (Cell Signaling; #2275), EEA1 (Clone C45B10; Cell Signaling; #3288), Rab13 (Sigma-Aldrich; #SAB4200058), and Human phospho-kinase array. Relative phosphorylation levels of 43 kinases and 2 LAMP1 (Clone D2D11; Cell Signaling; #9091S) were used at recommended dilu- related proteins were assessed using the Proteome Profiler Human Phospho-Kinase tions. Conjugated IRDye680RD/800CW secondary antibodies were employed and Array Kit (R&D Systems), according to the manufacturer’s instructions. In brief, detected using the Odyssey Infrared Imaging System (LI-COR Biosciences). cell lysates were incubated overnight with nitrocellulose membranes of the Human Quantification and analysis were performed using LI-COR Image Studio Software. Phospho-Kinase Array (R&D Systems). Membranes were then washed, incubated Experiments were repeated in triplicate. Unprocessed images of all western blots with biotinylated detection antibody cocktails, and then incubated with are shown in Supplementary Figs. 6 and 7. streptavidin-horseradish peroxidase and visualized using chemoluminescent reagents and X-ray film (GE Healthcare). Experiments were performed in four Live-cell microscopy. HEK293 (20,000 cells/well) and primary and immortalized biological replicates. The signal of each capture spot was measured using the Schwann cells (7500 cells/well) were seeded on top of glass slides with polystyrene ‘Protein Array Analyzer for ImageJ’ and normalized to internal reference controls. chambers (Lab-Tek II Chambered Coverglass slides; Nunc). Cells were incubated The median value of the four replicates was calculated for each target, and the fold for 20 min at 37 °C with pHrodo Red dextran 10,000 MW probes (ThermoFisher change compared to the control condition represented. Heteroscedastic Student’s t- Scientific; #P10361), at a concentration of 40 μg/ml, following the manufacturer’s test was used to determine statistical significance.

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y

Statistical analysis. Statistical analysis was performed using Prism 7 (GraphPad). 24. Kim, J. et al. MYBL1 rearrangements and MYB amplification in breast A P value of <0.05 was considered significant. Student’s t-test was employed for the adenoid cystic carcinomas lacking the MYB-NFIB fusion gene. J. Pathol. 244, comparison of means in parametric data. The heteroscedasticity was assessed for 143–150 (2018). each comparison, and homoscedastic or heteroscedastic t-tests were employed as 25. Gurzu, S. et al. The immunohistochemical profile of granular cell appropriate. All P values were two-sided. The 95% confidence intervals were (Abrikossoff) tumor suggests an endomesenchymal origin. Arch. Dermatol. adopted. Res. 307, 151–157 (2015). 26. Rejas, R. A., Campos, M. S., Cortes, A. R., Pinto, D. D. & de Sousa, S. C. The neural Data availability histogenetic origin of the oral granular cell tumor: an immunohistochemical evidence. Med. Oral. Patol. Oral. Cir. Bucal 16,e6–e10 (2011). Whole-exome sequencing and RNA sequencing data that support the findings of this 27. Ordonez, N. G. & Mackay, B. Granular cell tumor: a review of the pathology study have been deposited in Sequence Read Archive with the accession codes and histogenesis. Ultrastruct. Pathol. 23, 207–222 (1999). SRP119539 and SRP118840 respectively. 28. Kinouchi, K. et al. The (pro)renin receptor/ATP6AP2 is essential for vacuolar H+-ATPase assembly in murine cardiomyocytes. Circ. Res. 107,30–34 Received: 11 April 2018 Accepted: 2 August 2018 (2010). 29. Graham, L. A., Powell, B. & Stevens, T. H. Composition and assembly of the yeast vacuolar H(+)-ATPase complex. J. Exp. Biol. 203, 61-70 (2000). 30. Hirata, R., Umemoto, N., Ho, M. N., Ohya, Y., Stevens, T. H. & Anraku, Y. VMA12 is essential for assembly of the vacuolar H(+)-ATPase subunits onto the vacuolar membrane in Saccharomyces cerevisiae. J. Biol. Chem. 268, 961- References 967 (1993). 1. Machado, I., Cruz, J., Lavernia, J. & Llombart-Bosch, A. Solitary, multiple, 31. Sorkin, A. & von Zastrow, M. Endocytosis and signalling: intertwining – benign, atypical, or malignant: the “Granular Cell Tumor” puzzle. Virchows molecular networks. Nat. Rev. Mol. Cell Biol. 10, 609 622 (2009). Arch. 468, 527–538 (2016). 32. Fanburg-Smith, J. C., Meis-Kindblom, J. M., Fante, R. & Kindblom, L. G. 2. Nasser, H., Danforth, R. D. Jr., Sunbuli, M. & Dimitrijevic, O. Malignant Malignant granular cell tumor of soft tissue: diagnostic criteria and – granular cell tumor: case report with a novel karyotype and review of the clinicopathologic correlation. Am. J. Surg. Pathol. 22, 779 794 (1998). literature. Ann. Diagn. Pathol. 14, 273–278 (2010). 33. Shah, S. P. et al. Mutation of FOXL2 in granulosa-cell tumors of the ovary. – 3. Akahane, K. et al. Malignant granular cell tumor of the breast: case report and N. Engl. J. Med. 360, 2719 2729 (2009). literature review. Breast Cancer 22, 317–323 (2015). 34. Tognon, C. et al. Expression of the ETV6-NTRK3 gene fusion as a primary – 4. Ashworth, A., Lord, C. J. & Reis-Filho, J. S. Genetic interactions in cancer event in human secretory breast carcinoma. Cancer Cell. 2, 367 376 (2002). progression and treatment. Cell 145,30–38 (2011). 35. Piscuoglio, S. et al. Uterine adenosarcomas are mesenchymal neoplasms. – 5. Rivera, M. N. et al. An X chromosome gene, WTX, is commonly inactivated in J. Pathol. 238, 381 388 (2016). Wilms tumor. Science 315, 642–645 (2007). 36. Zhang, J. et al. INTEGRATE: gene fusion discovery using whole genome and – 6. Gao, J. et al. Integrative analysis of complex cancer genomics and clinical transcriptome data. Genome Res. 26, 108 118 (2015). profiles using the cBioPortal. Sci. Signal. 6, pl1 (2013). 37. McPherson, A. et al. deFuse: an algorithm for gene fusion discovery in tumor 7. Forgac, M. Vacuolar ATPases: rotary proton pumps in physiology and RNA-Seq data. PLoS Comput. Biol. 7, e1001138 (2011). pathophysiology. Nat. Rev. Mol. Cell Biol. 8, 917–929 (2007). 38. Cancer Genome Atlas Network. Comprehensive molecular portraits of human – 8. Louagie, E. et al. Role of furin in granular acidification in the endocrine breast tumours. Nature 490,61 70 (2012). pancreas: identification of the V-ATPase subunit Ac45 as a candidate 39. Shugay, M., Ortiz de Mendibil, I., Vizmanos, J. L. & Novo, F. J. Oncofuse: a substrate. Proc. Natl. Acad. Sci. USA 105, 12319–12324 (2008). computational framework for the prediction of the oncogenic potential of – 9. Kinouchi, K. et al. The role of individual domains and the significance of gene fusions. Bioinformatics 29, 2539 2546 (2013). fi shedding of ATP6AP2/(pro)renin receptor in vacuolar H(+)-ATPase 40. Martelotto, L. G. et al. Whole-genome single-cell copy number pro ling from fi fi – biogenesis. PLoS ONE 8, e78603 (2013). formalin- xed paraf n-embedded samples. Nat. Med. 23, 376 385 (2017). 10. Korvatska, O. et al. Altered splicing of ATP6AP2 causes X-linked 41. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows- – parkinsonism with spasticity (XPDS). Hum. Mol. Genet. 22, 3259–3268 Wheeler transform. Bioinformatics 25, 1754 1760 (2009). (2013). 42. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework 11. Gu, F., Aniento, F., Parton, R. G. & Gruenberg, J. Functional dissection of for analyzing next-generation DNA sequencing data. Genome Res. 20, – COP-I subunits in the biogenesis of multivesicular endosomes. J. Cell Biol. 1297 1303 (2010). 139, 1183–1195 (1997). 43. Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure – 12. Nishi, T. & Forgac, M. The vacuolar (H+)-ATPases--nature’s most versatile and heterogeneous cancer samples. Nat. Biotechnol. 31, 213 219 (2013). proton pumps. Nat. Rev. Mol. Cell Biol. 3,94–103 (2002). 44. Saunders, C. T. et al. Strelka: accurate somatic small-variant calling from – 13. Sigismund, S. et al. Endocytosis and signaling: cell logistics shape the sequenced tumor-normal sample pairs. Bioinformatics 28, 1811 1817 (2012). eukaryotic cell plan. Physiol. Rev. 92, 273–366 (2012). 45. Koboldt, D. C. et al. VarScan 2: somatic mutation and copy number alteration – 14. Bix, G. et al. Endorepellin causes endothelial cell disassembly of actin discovery in cancer by exome sequencing. Genome Res. 22, 568 576 (2012). cytoskeleton and focal adhesions through alpha2beta1 integrin. J. Cell Biol. 46. Ng, C. K. Y. et al. The Landscape of Somatic Genetic Alterations in 166,97–109 (2004). Metaplastic Breast Carcinomas. Clin. Cancer Res. 23, 3859-3870 (2017). 15. Liang, J. & Mills, G. B. AMPK: a contextual oncogene or tumor suppressor? 47. Thorvaldsdottir, H., Robinson, J. T. & Mesirov, J. P. Integrative Genomics Cancer Res. 73, 2929–2935 (2013). Viewer (IGV): high-performance genomics data visualization and exploration. – 16. Wu, D. & Pan, W. GSK3: a multifaceted kinase in Wnt signaling. Trends Brief. Bioinform. 14, 178 192 (2013). fi Biochem. Sci. 35, 161–168 (2010). 48. Shen, R. & Seshan, V. E. FACETS: allele-speci c copy number and clonal 17. Panek, R. L. et al. In vitro pharmacological characterization of PD 166285, a heterogeneity analysis tool for high-throughput DNA sequencing. Nucleic new nanomolar potent and broadly active protein tyrosine kinase inhibitor. Acids Res. 44, e131 (2016). fi J. Pharmacol. Exp. Ther.283, 1433-1444 (1997). 49. Carter, S. L. et al. Absolute quanti cation of somatic DNA alterations in – 18. Brandvold, K. R., Steffey, M. E., Fox, C. C. & Soellner, M. B. Development of a human cancer. Nat. Biotechnol. 30, 413 421 (2012). highly selective c-Src kinase inhibitor. ACS Chem. Biol. 7, 1393-1398 (2012). 50. Landau, D. A. et al. Evolution and impact of subclonal mutations in chronic – 19. Muller, J., Sperl, B., Reindl, W., Kiessling, A. & Berg, T. Discovery of lymphocytic leukemia. Cell 152, 714 726 (2013). chromone-based inhibitors of the transcription factor STAT5. Chembiochem 51. Chang, M. T. et al. Identifying recurrent mutations in cancer reveals fi 9, 723–727 (2008). widespread lineage diversity and mutational speci city. Nat. Biotechnol. 34, – 20. Jansen, E. J. et al. ATP6AP1 deficiency causes an immunodeficiency with 155 163 (2016). hepatopathy, cognitive impairment and abnormal protein glycosylation. Nat. 52. Rimmer, A. et al. Integrating mapping-, assembly- and haplotype-based Commun. 7, 11600 (2016). approaches for calling variants in clinical sequencing applications. Nat. Genet. – 21. Hedera, P., Alvarado, D., Beydoun, A. & Fink, J. K. Novel mental retardation- 46, 912 918 (2014). epilepsy syndrome linked to Xp21.1-p11.4. Ann. Neurol. 51,45–50 (2002). 53. Narzisi, G. et al. Accurate de novo and transmitted indel detection in exome- – 22. Ramser, J. et al. A unique exonic splice enhancer mutation in a family with X- capture data using microassembly. Nat. Methods 11, 1033 1036 (2014). fi linked mental retardation and epilepsy points to a novel role of the renin 54. Uchida, T. et al. Clonality analysis by methylation-speci c PCR for the human – receptor. Hum. Mol. Genet. 14, 1019–1027 (2005). androgen-receptor gene (HUMARA-MSP). Leukemia 14, 207 212 (2000). 23. Weinreb, I. et al. Hotspot activating PRKD1 somatic mutations in 55. Busque, L. et al. An expression based clonality assay at the human androgen – polymorphous low-grade adenocarcinomas of the salivary glands. Nat. Genet. receptor locus (HUMARA) on chromosome X. Nucleic Acids Res. 22, 697 698 46, 1166–1169 (2014). (1994).

12 NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05886-y ARTICLE

56. Boudewijns, M., van Dongen, J. J. & Langerak, A. W. The human androgen Author contributions receptor X-chromosome inactivation assay for clonality diagnostics of natural B.P.R.,B.W.andJ.S.R.-F.conceivedofandsupervisedthestudy.E.A.R.,I.O.E,M.I.A. killer cell proliferations. J. Mol. Diagn. 9, 337–344 (2007). E, J.P., T.H., M.E. and B.P.R. provided samples. P.S., R.K., D.N.B., R.G.-M., R.S.L., 57. Leiserson, M. D., Wu, H. T., Vandin, F. & Raphael, B. J. CoMEt: a statistical I.d.B. and P.B. performed bioinformatics analyses.F.P.,F.C.G.,T.H.andM.E.cen- approach to identify combinations of mutually exclusive alterations in cancer. trally reviewed cases. Sample processing and Sanger sequencing were performed by Genome Biol. 16, 160 (2015). F.P., F.C.G., A.D.C.P., B.A. and R.B. Functional experiments were carried out by F.P., 58. Riaz, N. et al. Pan-cancer analysis of bi-allelic alterations in homologous A.B., T.B., D.F., A.D.C.P. and S.F. Data acquisition, interpretation, and analysis recombination DNA repair genes. Nat. Commun. 8, 857 (2017). were performed by F.P., A.B., T.B., P.S., F.C.G., D.F., A.D.C.P., R.K., D.N.B., R.G.-M., 59. McWilliam, H. et al. Analysis Tool Web Services from the EMBL-EBI. Nucleic B.A., R.B., R.S.L., I.d.B., S.F., R.G., E.F., A.L., E.M.d.S., J.R.L., P.B., L.C.G., A.A.J., L.N., Acids Res. 41, W597–W600 (2013) B.P.R., B.W. and J.S.R.-F. The manuscript was initially drafted by F.P., A.B., T.B., B.P. 60. Jelinic, P. et al. Recurrent SMARCA4 mutations in small cell carcinoma of the R.,B.W.andJ.S.R.-F.andallauthorseditedandapprovedthefinal draft of the ovary. Nat. Genet. 46, 424–426 (2014). manuscript. 61. Brough, R. et al. APRIN is a cell cycle specific BRCA2-interacting protein required for genome integrity and a predictor of outcome after chemotherapy Additional information in breast cancer. EMBO J. 31, 1160–1176 (2012). 62. Chiang, S. et al. IDH2 mutations define a unique subtype of breast cancer with Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- altered nuclear polarity. Cancer Res. 76, 7118–7129 (2016). 018-05886-y. 63. Li, H., Chang, L. J., Neubauer, D. R., Muir, D. F. & Wallace, M. R. Immortalization of human normal and NF1 neurofibroma Schwann cells. Lab. Competing interests: The authors declare no competing interests. Invest. 96, 1105–1115 (2016). 64. Martelotto, L. G. et al. Genomic landscape of adenoid cystic carcinoma of the Reprints and permission information is available online at http://npg.nature.com/ breast. J. Pathol. 237, 179–189 (2015). reprintsandpermissions/ 65. Fellmann, C. et al. An optimized microRNA backbone for effective single- – Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in copy RNAi. Cell Rep. 5, 1704 1713 (2013). fi 66. Chen, H. J. et al. A recellularized human colon model identifies cancer driver published maps and institutional af liations. genes. Nat. Biotechnol. 34, 845–851 (2016). 67. Weigelt, B. Warne P. H. & Downward J. PIK3CA mutation, but not PTEN loss of function, determines the sensitivity of breast cancer cells to mTOR inhibitory drugs. Oncogene 30, 3222–3233 (2011). Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Acknowledgements Commons license, and indicate if changes were made. The images or other third party ’ Research reported in this publication was funded in part by the Functional Genomics material in this article are included in the article s Creative Commons license, unless Initiative, and in part by a Cancer Center Support Grant of the National Institutes of indicated otherwise in a credit line to the material. If material is not included in the ’ Health/National Cancer Institute (grant no. P30CA008748). J.S.R.-F. is funded in part by article s Creative Commons license and your intended use is not permitted by statutory the Breast Cancer Research Foundation. B.A. was funded by CAPES (#PDSE- regulation or exceeds the permitted use, you will need to obtain permission directly from 88881.133984/2016-01). The transmission electron microscope at the Imaging Core/ the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Electron Microscopy & Histology Services of Weill Cornell Medicine was obtained with licenses/by/4.0/. an NIH Shared Instrumentation Grant for Shared Resources (S10RR027699). The con- tent is solely the responsibility of the authors and does not necessarily represent the © The Author(s) 2018 official views of the National Institutes of Health.

NATURE COMMUNICATIONS | (2018) 9:3533 | DOI: 10.1038/s41467-018-05886-y | www.nature.com/naturecommunications 13